As emotions play a central role in human communication, automatic emotion recognition has attracted increasing attention in the last two decades. While multimodal systems enjoy high performances on lab-controlled data, they are still far from providing ecological validity on non-lab-controlled, namely “in-the-wild” data. This work investigates audiovisual deep learning approaches to emotion recognition in in-the-wild problem. Inspired by the outstanding performance of end-to-end and transfer learning techniques, we explored the effectiveness of architectures in which a modality-specific Convolutional Neural Network (CNN) is followed by a Long Short-Term Memory Recurrent Neural Network (LSTM-RNN) using the AffWild2 dataset under the Affective Behavior Analysis in-the-Wild (ABAW) challenge protocol. We deployed unimodal end-to-end and transfer learning approaches within a multimodal fusion system, which generated final predictions using a weighted score fusion scheme. Exploiting the proposed deep-learning-based multimodal system, we reached a test set challenge performance measure of 48.1% on the ABAW 2020 Facial Expressions challenge, which advances the first-runner-up performance.

The article presents an analytical review of research in the affective computing field. This research direction is a component of artificial intelligence, and it studies methods, algorithms and systems for analyzing human affective states during interactions with other people, computer systems or robots. In the field of data mining, the definition of affect means the manifestation of psychological reactions to an exciting event, which can occur both in the short and long term, and also have different intensity. The affects in this field are divided into 4 types: affective emotions, basic emotions, sentiment and affective disorders. The manifestation of affective states is reflected in verbal data and non-verbal characteristics of behavior: acoustic and linguistic characteristics of speech, facial expressions, gestures and postures of a person. The review provides a comparative analysis of the existing infoware for automatic recognition of a person’s affective states on the example of emotions, sentiment, aggression and depression. The few Russian-language, affective databases are still significantly inferior in volume and quality compared to electronic resources in other world languages. Thus, there is a need to consider a wide range of additional approaches, methods and algorithms used in a limited amount of training and testing data, and set the task of developing new approaches to data augmentation, transferring model learning and adapting foreign-language resources. The article describes the methods of analyzing unimodal visual, acoustic and linguistic information, as well as multimodal approaches for the affective states recognition. A multimodal approach to the automatic affective states analysis makes it possible to increase the accuracy of recognition of the phenomena compared to single-modal solutions. The review notes the trend of modern research that neural network methods are gradually replacing classical deterministic methods through better quality of state recognition and fast processing of large amount of data. The article discusses the methods for affective states analysis. The advantage of multitasking hierarchical approaches is the ability to extract new types of knowledge, including the influence, correlation and interaction of several affective states on each other, which potentially leads to improved recognition quality. The potential requirements for the developed systems for affective states analysis and the main directions of further research are given.

Recognition of emotions and sentiment (affective states) from human audio–visual information is widely used in healthcare, education, entertainment, and other fields; therefore, it has become a highly active research area. The large variety of corpora with heterogeneous data available for the development of single-corpus approaches for recognition of affective states may lead to approaches trained on one corpus being less effective on another. In this article, we propose a multi-corpus learned audio–visual approach for emotion and sentiment recognition. It is based on the extraction of mid-level features at the segment level using two multi-corpus temporal models (a pretrained transformer with GRU layers for the audio modality and pre-trained 3D CNN with BiLSTM-Former for the video modality) and on predicting affective states using two single-corpus cross-modal gated self-attention fusion (CMGSAF) models. The proposed approach was tested on the RAMAS and CMU-MOSEI corpora. To date, our approach has outperformed state-of-the-art audio–visual approaches for emotion recognition by 18.2% (78.1% vs. 59.9%) for the CMU-MOSEI corpus in terms of the Weighted Accuracy and by 0.7% (82.8% vs. 82.1%) for the RAMAS corpus in terms of the Unweighted Average Recall.

In this paper we present a novel method designed for word-level visual speech recognition and intended for use in human-robot interaction. The ability of robots In this paper we present a novel method designed for word-level visual speech recognition and intended for use in human-robot interaction. The ability of robots